A deep learning system for detecting diabetic retinopathy across the disease spectrum

TLDR

Retinal screening enables early detection and timely treatment of diabetic retinopathy. The study develops DeepDR, a deep learning system to detect all stages of diabetic retinopathy. DeepDR was trained on 466,247 fundus images from 121,342 diabetic patients and evaluated on 200,136 images from 52,004 patients plus three external datasets totaling 209,322 images. The system achieved AUCs of 0.901–0.967 for lesion detection and 0.943–0.972 for grading, with external validation AUCs ranging from 0.916 to 0.970, confirming its effectiveness.

Abstract

Retinal screening contributes to early detection of diabetic retinopathy and timely treatment. To facilitate the screening process, we develop a deep learning system, named DeepDR, that can detect early-to-late stages of diabetic retinopathy. DeepDR is trained for real-time image quality assessment, lesion detection and grading using 466,247 fundus images from 121,342 patients with diabetes. Evaluation is performed on a local dataset with 200,136 fundus images from 52,004 patients and three external datasets with a total of 209,322 images. The area under the receiver operating characteristic curves for detecting microaneurysms, cotton-wool spots, hard exudates and hemorrhages are 0.901, 0.941, 0.954 and 0.967, respectively. The grading of diabetic retinopathy as mild, moderate, severe and proliferative achieves area under the curves of 0.943, 0.955, 0.960 and 0.972, respectively. In external validations, the area under the curves for grading range from 0.916 to 0.970, which further supports the system is efficient for diabetic retinopathy grading.